LLM & Agents

This skill should be used when conducting comprehensive research on any topic using the OpenAI Deep Research API. It automates prompt enhancement through interactive clarifying questions, saves research parameters, and executes deep research with web search capabilities. Use when the user asks for in-depth analysis, investigation, research summaries, or topic exploration.

Create specialized Claude Code agents (subagents) with proper YAML frontmatter, tool permissions, and domain expertise. Use when: creating new agents, designing agent prompts, configuring agent tools and models. Triggers: create agent, new agent, agent template, subagent, agent design.

Validates and reviews Claude Code agents against best practices. Checks YAML syntax, naming conventions, description quality, system prompt effectiveness, and tool configuration. Provides improvement suggestions and optionally applies fixes. Use when validating agents, reviewing agent quality, checking agents before commit, or when `--validate-agent` is mentioned.

Use when pair programming with humans - enforces XP values (communication, simplicity, feedback, courage, respect) to deliver high-quality software; push back on YAGNI violations regardless of seniority or sunk cost

Claude Code hooks development guide covering all 10 hook events lifecycle, PostToolUse visibility patterns, PreToolUse guards, Stop hook schema, and debugging. Use when creating hooks, troubleshooting hook output, understanding hook lifecycle, or when user mentions decision block, hook JSON output, stop hook, or Claude Code hooks.

Add a new AI provider or model for recipe generation. Use when adding support for a new LLM provider (Anthropic, Google, etc.) or adding models to an existing provider.

Enforce line limits (CLAUDE.md ≤100, others ≤200) and remove low-ROI content from .md files, use when optimizing or refactoring docs (project)

智能计量经济学分析代理。当用户输入以"autoregmonkey："开头时，LLM会解析经济学计量任务，参考RAG数据库知识，动态调用Python和Stata技能执行任务，最后生成中文报告。

Creates and manages plugin marketplaces for Claude Code, enabling plugin discovery, distribution, and team sharing. Generates marketplace.json files with proper schema, configures plugin sources (GitHub, Git, local), and sets up team configurations for automatic marketplace installation. Use when publishing Claude Code plugins, creating internal plugin catalogs, setting up team plugin distribution, or managing plugin discovery for Claude Code projects.

Pragmatic Rust conventions to keep code readable, testable, and performant for this project.

Letta framework for building stateful AI agents with long-term memory. Use for AI agent development, memory management, tool integration, and multi-agent systems.

Run a comprehensive code quality review. Checks security, performance, maintainability, project conventions (CLAUDE.md), architecture, and testing. Runs rustfmt and bazel test. Use /quality to check code quality or validate before committing.

Router for systems thinking methodology - patterns, leverage points, archetypes, stocks-flows, causal loops, BOT graphs

This skill should be used when the model's ROLE_TYPE is orchestrator and needs to delegate tasks to specialist sub-agents. Provides scientific delegation framework ensuring world-building context (WHERE, WHAT, WHY) while preserving agent autonomy in implementation decisions (HOW). Use when planning task delegation, structuring sub-agent prompts, or coordinating multi-agent workflows.

Multi-model AI council for actionable project advice. Leverages Gemini 3 Flash and GPT-5.2 skills in parallel, then synthesizes through an Opus Judge for stage-appropriate, non-overkill recommendations. Use when seeking architectural guidance, code review synthesis, or implementation planning.

Validate markdown link portability in Claude Code skills and plugins. Use when checking links, validating portability, fixing broken links, absolute paths, relative paths, or before skill distribution. Ensures relative paths for cross-installation compatibility.

When calling LLM APIs from Python code. When connecting to llamafile or local LLM servers. When switching between OpenAI/Anthropic/local providers. When implementing retry/fallback logic for LLM calls. When code imports litellm or uses completion() patterns.

Apply Rules Repository conventions and forward-only workflow. Use when generating or editing docs, prompts, or code in projects that consume this repository or when this repository is the active workspace. Ensure Pact → Rules → Guides → Implementation links are created and updated, and prefer linking to submodule guides over duplicating content.

This skill should be used when teams need to adopt Supertester to build deterministic Elixir OTP tests with isolation, synchronization, and supervision coverage.

Creates specialized subagents for Claude Code with proper configuration, capabilities, tool restrictions, and Task tool integration. Use when building specialized agents, editing agent files, creating subagents for specific tasks, or when users mention agent development, subagent configuration, or Task tool usage. Different from Skills - agents are invoked via Task tool.